Uncertainty Estimate of Surface Irradiances Computed with MODIS-, CALIPSO-, and CloudSat-Derived Cloud and Aerosol Properties

Seiji Kato 0 1 Norman G. Loeb 0 1 David A. Rutan 0 1 Fred G. Rose 0 1 Sunny Sun-Mack 0 1 Walter F. Miller 0 1 Yan Chen 0 1 0 D. A. Rutan F. G. Rose S. Sun-Mack W. F. Miller Y. Chen Science System & Applications Inc , Hampton, VA, USA 1 S. Kato (&) N. G. Loeb Climate Science Branch, NASA Langley Research Center , Hampton, VA 23681-2199, USA Differences of modeled surface upward and downward longwave and shortwave irradiances are calculated using modeled irradiance computed with active sensor-derived and passive sensor-derived cloud and aerosol properties. The irradiance differences are calculated for various temporal and spatial scales, monthly gridded, monthly zonal, monthly global, and annual global. Using the irradiance differences, the uncertainty of surface irradiances is estimated. The uncertainty (1r) of the annual global surface downward longwave and shortwave is, respectively, 7 W m-2 (out of 345 W m-2) and 4 W m-2 (out of 192 W m-2), after known bias errors are removed. Similarly, the uncertainty of the annual global surface upward longwave and shortwave is, respectively, 3 W m-2 (out of 398 W m-2) and 3 W m-2 (out of 23 W m-2). The uncertainty is for modeled irradiances computed using cloud properties derived from imagers on a sun-synchronous orbit that covers the globe every day (e.g., moderate-resolution imaging spectrometer) or modeled irradiances computed for nadir view only active sensors on a sun-synchronous orbit such as Cloud-Aerosol Lidar and Infrared Pathfinder Satellite Observation and CloudSat. If we assume that longwave and shortwave uncertainties are independent of each other, but up- and downward components are correlated with each other, the uncertainty in global annual mean net surface irradiance is 12 W m-2. One-sigma uncertainty bounds of the satellite-based net surface irradiance are 106 W m-2 and 130 W m-2. Estimating the surface irradiance is important in understanding the energy cycle of the globe for several reasons. The sum of surface net irradiance and other surface enthalpy - (sensible and latent heat) fluxes is the energy flux through the lower boundary of the atmospheric column and the upper boundary of an ocean column. Therefore, the global mean net surface irradiance balances with the sum of the surface latent and sensible heat fluxes and ocean heating rate (Wong et al. 2006). In addition, the radiative net energy deposition in the atmosphere and vertical and horizontal profiles of the energy deposition determine the dynamics in the atmosphere. Understanding the top-of-atmosphere (TOA) surface and atmospheric irradiances quantitatively is, therefore, necessary to quantitatively understand the dynamics, which in turn controls cloud feedback processes (Wielicki et al. 1995). The global mean surface irradiance estimate is only possible through modeling surface irradiances. In earlier studies, cloud properties derived from passive satellite instruments combined with radiative transfer models have been used to estimate surface irradiance (e.g., Pinker and Laszlo 1992; Zhang et al. 1995, and a summary is given by Kandel and Viollier 2010). While surface irradiances computed with satellite-derived and modeled cloud properties have been compared in earlier studies (e.g., Hatzianastassiou et al. 2005; Su et al. 2008; Stephens 2011), the uncertainty of surface irradiances averaged over different temporal and spatial scales, such as monthly or annual, and regional, zonal, or global, is not well understood. The purpose of this paper is to extend the study by Kato et al. (2011) to estimate uncertainties of surface irradiance components in various spatial and temporal scales (1 9 30 or 1 9 1 gridded, 1 zonal, and global spatial scales and monthly and annual temporal scales). In this study, we define the uncertainty as a range of surface irradiances in which the true value resides at a 68% probability. Our goal is different from estimating the error of a specific surface irradiance estimate, although we need to have a specific surface irradiance estimate to attach the uncertainty. Taylor and Kuyatt (1994) describe the difference nicely; the result of a measurement (modeled irradiance) could have a negligible error because it can unknowably be very close to the truth even though it may have a large uncertainty. There are two possible ways of estimating the modeled surface irradiance uncertainty. One way is to estimate the uncertainty of the input variables, perturb inputs by the uncertainty amount, and compute the irradiance using a radiative transfer model. The irradiance change by the sensitivity study is considered as the uncertainty due to input variables. Some earlier studies (e.g., Zhang et al. 1995; Zhang et al. 2004; Kim and Ramanathan 2008) used this approach. A second approach is to use surface observations and compute the root mean square (RMS) difference of modeled and observed surface irradiances. We primarily take the former approach in this study, but briefly examine how uncertainties derived by the two approaches differ. Sections 2 and 3 present a brief overview of surface longwave and shortwave irradiance estimates, respectively, to understand the range of the global annual mean values. Section 4 briefly discusses the computation method using CALIPSO (Winker et al. 2010), CloudSat (Stephens et al. 2008), moderate resolution imaging spectrometer (MODIS) and Clouds and the Earths Radiant Energy System (CERES) data. Section 5 analyzes the uncertainty of surface irradiances for various temporal and spatial scales by comparing two modeled irradiances (sensitivity approach). Section 6 uses surface observations to estimate modeled surface irradiance uncertainties (surface observation approach). Section 7 combines uncertainties of all surface irradiance components and discusses the uncertainty in the global annual net surface irradiance. 2 Global Annual Mean Surface Downward Longwave Irradiance Stephens et al. (2011) provide a summary of the global annual mean surface longwave upward and downward irradiance estimated from satellite observations, reanalysis, and ground-based observations. The global annual mean downward longwave irradiance estimated from satellite observations (GEWEX SRB, ISCCP-FD, CERES) ranges from 342 to 348 W m-2 (Stephens et al. 2011). The global annual mean downward longwave irradiance estimated by reanalyses ranges from 324 to 340 W m-2 (Stephens et al. 2011). Wild et al. (2001) compared the global annual mean surface downward longwave irradiance computed in general circulation models (GCMs) with surface observations in Global Energy Balance Archive (GEBA) and Baseline Surface Radiation Network (BSRN, Ohmura et al. 1998) data sets. Their results show that the modeled global annual mean surface downward longwave irradiance by GCMs varies more than 40 W m-2, ranging from 303 to 344 W m-2. GCM-derived surface downward longwave irradiances are less than observed irradiances especially under dry and cold co (...truncated)